A Georgia Tech and Penn State team tested a NAND flash variant that stores data via a crystalline ferroelectric material whose built in electric polarity, set to one of two stable states, holds each bit; the design tolerated roughly 1 million rads of radiation, a dose far beyond…
Deep-space missions are on track to outgrow the memory chips that carry their data home. A Georgia Tech and Penn State team is now pointing at one possible way out: a NAND flash variant that stores information using the natural polarity of a ferroelectric material, rather than the trapped electrical charge that conventional flash relies on. In testing, the design tolerated about 1 million rads of radiation, roughly 30 times what standard NAND endures under the same conditions, according to the team's paper in Nano Letters.
For context, deep-space and cislunar missions generate more scientific data than older storage can survive to bring home, and radiation tolerance has been a quiet constraint on what missions can keep. Asif Khan, an electrical and computer engineering researcher at Georgia Tech, said the design "could expand the range of missions that rely on solid-state data storage," a position released through the university's Research Horizons write-up on the result.
The mechanism is the headline detail. Conventional flash memory stores bits by holding or releasing electrical charge in a floating gate; high-energy particles knock that charge around and corrupt the data. Ferroelectric materials, which hold a built-in electric polarity that flips between two stable states, sidestep that failure mode. The team's design layers a ferroelectric film onto the storage cell, and the polarity of that film is what carries the bit.
This is a research result, not a shipping product. No foundry has qualified the design, no mission has adopted it, and the 1 million rad figure is a tested tolerance threshold under specific conditions, not a guaranteed spec for every deep-space mission profile. Treat the 30x advantage as a relative comparison to conventional NAND in the same test, not as an absolute ceiling.
The same Research Bits: June 15 roundup, curated by Jesse Allen for Semiconductor Engineering, surfaces two other recent notes on devices that handle information in non-standard ways, useful context for the broader question of how future electronics might keep working in extreme conditions.
In Montreal, a Polytechnique team integrated an organic molecule called triphenylamine-dicyanoquinoxaline, or TPA-QCN, onto silicon using vacuum evaporation. The result is a chip that can amplify, modulate, and convert infrared light to visible red on the same device, functions that usually require separate components. Stéphane Kéna-Cohen, an engineering physics professor at Polytechnique Montréal, said spontaneous molecular alignment during evaporation is the trick that makes on-chip second-order nonlinear optics practical. The work appears as "Poling-free integrated second-order nonlinear optics with evaporated organic thin films" in Science Advances 12, eaeg3170 (2026), and the university's news release frames it as a step toward photonic circuits that share silicon's manufacturing base.
Separately, a Seoul National University, Stanford, and Chinese Academy of Sciences collaboration built an electrochemical organic light-emitting transistor that runs on under 3.5 volts and accumulates light output when stimulated repeatedly, a memory-like behavior the team describes as neuromorphic. In a flexible wearable display demo, two 1.5-volt batteries were enough to drive it. Tae-Woo Lee, a materials science and engineering professor at SNU, framed the device as a single platform for on-skin displays, intelligent artificial skin, and wearable healthcare. The paper, "Ultralow-voltage electrochemical organic light-emitting transistors with pinned and wide lateral recombination," is in Nature Materials (2026).
The common thread across all three notes is information handling through mechanisms such as polarization, photonic functions, and light-emitting organic layers with memory, all of which go beyond what standard trapped-charge silicon memory does.
The next checkpoint for the ferroelectric NAND work is independent confirmation. The team has shown that the polarization-based cell survives the radiation dose. The open question is whether the design can be integrated into a real chip fab process at density, cost, and yield competitive with conventional flash, and whether the tolerance holds up across a broader range of mission profiles. The 1 million rad result is a credible research milestone. The path from there to a qualified memory in a spacecraft is the part that still needs work.